US20060165311A1  Spatial standard observer  Google Patents
Spatial standard observer Download PDFInfo
 Publication number
 US20060165311A1 US20060165311A1 US11045041 US4504105A US2006165311A1 US 20060165311 A1 US20060165311 A1 US 20060165311A1 US 11045041 US11045041 US 11045041 US 4504105 A US4504105 A US 4504105A US 2006165311 A1 US2006165311 A1 US 2006165311A1
 Authority
 US
 Grant status
 Application
 Patent type
 Prior art keywords
 image
 reference
 luminance
 test
 eq
 Prior art date
 Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
 Granted
Links
Images
Classifications

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T7/00—Image analysis
 G06T7/0002—Inspection of images, e.g. flaw detection

 H—ELECTRICITY
 H04—ELECTRIC COMMUNICATION TECHNIQUE
 H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
 H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
 H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding

 H—ELECTRICITY
 H04—ELECTRIC COMMUNICATION TECHNIQUE
 H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
 H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
 H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using preprocessing or postprocessing specially adapted for video compression
 H04N19/86—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using preprocessing or postprocessing specially adapted for video compression involving reduction of coding artifacts, e.g. of blockiness

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T2207/00—Indexing scheme for image analysis or image enhancement
 G06T2207/30—Subject of image; Context of image processing
 G06T2207/30168—Image quality inspection

 H—ELECTRICITY
 H04—ELECTRIC COMMUNICATION TECHNIQUE
 H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
 H04N17/00—Diagnosis, testing or measuring for television systems or their details
 H04N17/04—Diagnosis, testing or measuring for television systems or their details for receivers
Abstract
The present invention relates to devices and methods for the measurement and/or for the specification of the perceptual intensity of a visual image, or the perceptual distance between a pair of images. Grayscale test and reference images are processed to produce test and reference luminance images. A luminance filter function is convolved with the reference luminance image to produce a local mean luminance reference image. Test and reference contrast images are produced from the local mean luminance reference image and the test and reference luminance images respectively, followed by application of a contrast sensitivity filter. The resulting images are combined according to mathematical prescriptions to produce a Just Noticeable Difference, JND value, indicative of a Spatial Standard Observer, SSO. Some embodiments include masking functions, window functions, special treatment for images lying on or near borders and preprocessing of test images.
Description
 [0001]The invention described herein was made by employees of the United States Government and may be manufactured and used by or for the Government for governmental purposes without payment of any royalties thereon or therefor.
 [0002]3.a. Technical Field of the Invention
 [0003]This invention relates generally to the field of devices and methods for the specification and measurement of the perceptual intensity of one or more visual images and, more particularly, to the rapid and efficient determination of a visibility metric for such images.
 [0004]3.b. Description of the Prior Art
 [0005]Vision is the means by which most people acquire and process information about the world around them. Numerous objects intended for human use include a component of information to be identified visually by a human observer. Some everyday examples include information displayed on a screen or page, keys or buttons to be pressed on a keyboard, telephone, calculator, remote control unit, among many other examples. Therefore, it is reasonable that the design of such objects include specifications to insure that the visual information is accessible to typical human observers, that is, that the information is visible. Providing a means for measuring and specifying visibility, a “visibility metric,” is an objective of the present invention.
 [0006]A significant challenge in designing standards for visibility is that such standards are based upon models of the human visual sense. However, vision is a complex and only partially understood process. Previous standards for visibility have thus tended to be complex, difficult to use and not sufficiently general to serve as a standard method or methods for the specification and measurement of visibility. The performance of various visibility metrics has been reviewed by Ahumada and coworkers in two publications: Society for Information Display, International Symposium, Digest of Technical Papers Vol. 24, pp. 305308 (1993) and Vol. 26. pp. 4548 (1995), the contents of both publications are incorporated herein by reference.
 [0007]Other examples of visibility metrics include the work of Lubin and coworkers U.S. Pat. No. 6,654,504, US Patent Application Publication 2002/0031277 and “A Human System Model for Objective Picture Quality Measurements,” Proceedings, International Broadcasters'Convention, Amsterdam, The Netherlands, pp. 498503 (1997). These methods developed by Lubin and coworkers require extensive calibration for each application in addition to suffering from the disadvantage of complexity. These methods are chiefly intended for image quality evaluation.
 [0008]Other methods for estimating visibility include those of Barten, “The SQRI Method: A New Method for the Evaluation of Visible Resolution on a Display,” Proceedings of the Society for Information Display, Vol. 28, pp. 253262 (1987). In addition to complexity, the Barten method suffers from the further disadvantage of being appropriate primarily for the specification of displays such as television monitors.
 [0009]Standards for the measurement and specification of color are known in the art and widely used. However, such color standards typically do not address the spatial pattern employed in a visual signal (for example, the shape of a letter). Consequently, such methods are not appropriate for specifying or measuring visibility.
 [0010]Thus, a need exists in the art for a standard specification and measurement of visibility, sufficiently general to be applicable to large classes of visual information but sufficiently simple for widespread implementation and use, including embedding into inexpensive systems.
 [0011]Accordingly and advantageously, the present invention relates to systems and techniques for processing visual information to produce a single numerical value for the visibility metric indicative of a Spatial Standard Observer (SSO). Advantages of the SSO include a simple and efficient design that produces an accurate visibility metric with a relatively few calculations.
 [0012]Some embodiments of the present invention use a Minkowski sum directly over filtered image pixels. This technique avoids the need for complicated spatial frequency filter banks, with a corresponding gain in simplicity and computational efficiency.
 [0013]A particular form of Contrast Sensitivity Filter (CSF) is used in some embodiments of the present invention which combines radial and obliqueeffect filters. This permits accurate visibility predictions of the visibility of oblique patterns such as halftoning and rasterizing artifacts.
 [0014]Viewing distance and image resolution are jointly treated in an advantageous manner in some embodiments of the present invention. The use of this feature causes the computed value of image visibility to be substantially independent of image resolution (except to the extent that the resolution actually alters the visibility of the information in the image).
 [0015]A window function is advantageously employed in some embodiments of the present invention in such a manner as to represent the reduction in visibility with distance from the observer's region of fixation.
 [0016]It is advantageous in some embodiments of the present invention to use convolution operations along with the window function. In this manner it is feasible to simulate the scanning of an image by the eye of the observer.
 [0017]Pooling the data accumulates the visibility over the scan and is advantageously employed in some embodiments of the present invention.
 [0018]When images are located near a border region, it may occur that the border has a markedly different intensity (typically darker) than that of the image and the general image background. In such cases, it is advantageous in some embodiments of the present invention to introduce special procedures for handling border effects. Two examples are presented. One includes at least a portion of the border into the definition of “image” leading to a enhanced image that is then processed by the SSO. Another approach is to attenuate the image contrast near the border.
 [0019]The SSO provides a standardized measure of visibility, allowing comparisons to be made of visibility measurements taken in a wide variety of applications, locations and times. Manufacturing and engineering specifications of visibility in standardized units can then be made.
 [0020]Furthermore, SSO visibility measurements are not limited by target size. Thus, very large or very small displays can use SSO.
 [0021]The SSO further provides the feasibility of making simple, automated measurements of the visibility of visual information, not requiring the use of human observers to estimate visibility. Simplicity of measurement is an important feature of SSO in order to allow SSO to be adopted in a wide variety of applications and at low cost.
 [0022]SSO has numerous potential areas of application. We note a few applications as illustrative of the utility of SSO, not thereby limiting the scope of SSO to only those enumerated. Many other applications are apparent to those with ordinary skill in the art, within the scope of the present invention. Possible applications include:

 Photometric Instruments incorporating SSO to produce a “spatial photometer” for the measurement of the visibility of spatial patterns.
 Imaging Devices and Systems employing SSO to calculate the visibility of targets as viewed through those systems such as infrared viewing systems and remote viewing systems (e.g., as in unmanned aerial vehicles).
 Copier Manufacturing employing SSO to measure the visibility of defects produced by copiers and thus test the copier and/or improve copier design.
 Video Codecs employing SSO in testing and/or design to measure the visibility of image compression artifacts with a view towards reducing visible defects and increasing bitrate.
 Display Manufacturing employing SSO to detect and measure visible artifacts with a view towards improving and automating product quality control and output by only rejecting devices having visible artifacts.
 Graphics Software employing SSO to estimate the visibility of graphic elements and/or to estimate the visibility of artifacts due to the rendering process.
 Predicting Visual Performance of Humans Following Vision Correction using SSO and thereby preevaluate the relative efficacy of various correction procedures before surgery.
 Digital Watermarking employing SSO to calculate the visibility of a recoverable signature labeling an image that is intended to be invisible to a human viewer.

 [0031]These are among the advantages achieved in accordance with various embodiments of the present invention as described in detail below.
 [0032]To facilitate understanding, identical reference numerals have been used, where possible, to designate identical elements that are common to the figures.
 [0033]The techniques of the present invention can readily be understood by considering the following detailed description in conjunction with the following drawings, in which:
 [0034]
FIG. 1 depicts a highlevel block diagram of a typical embodiment of a Spatial Standard Observer used to compute a visibility metric.  [0035]
FIG. 2 depicts a typical target adjacent to a dark border surrounding the image.  [0036]
FIG. 3 depicts a typical border aperture function that, in this example, has 240 columns, 180 rows, each pixel is ( 1/60) degree in height and width. The value of the parameter bscale is 0.50 deg., and bgain is 1.  [0037]
FIG. 4 depicts a highlevel block diagram of an exemplary computer system that can be used for implementation of techniques of the Spatial Standard Observer.  [0038]After considering the following description, those skilled in the art will clearly realize that the teachings of the invention can be readily utilized for determining the probable visibility of various graphical or visual depictions and displays as viewed by a typical human observer. In particular, the present invention relates generally to systems and techniques for processing one or more images to produce a single numerical value, or “visibility metric,” indicative of a “Spatial Standard Observer” (SSO). Advantages of the present invention include techniques for the rapid evaluation of the SSO.
 [0039]The present invention relates generally to devices and methods for the measurement and/or for the specification of the perceptual intensity of a visual image. Other embodiments relate generally to devices and methods for the measurement and/or for the specification of differences in perception or “perceptual distance” between two or more visual images. Such devices and methods can be advantageously used in situations in which it is desired to measure or to specify visibility or visual intensity. Examples include the determination of visibility and/or discriminability of text, graphic elements, labels, icons, among other visual images. Examples also include the determination of visibility and/or discriminability between images, such as an original image and a compressed digital form of that image. Some embodiments of the present invention can also be advantageously used to quantify the visibility of blemishes on a display as might be useful, for example, in providing objective determinations of pass/fail criteria in the manufacture of displays.
 [0040]In essence, various embodiments of the present invention operate on a digital image (or an analog image following digitization) or on a pair of digital images. An arbitrary number of images can be compared by repeated pairwise comparisons. Thus, for economy of language we will describe applications of the present invention to a single digital image or to the comparison of two digital images, understanding that this is by way of illustration and not limitation since multiple images can be handled by multiple applications of such pairwise comparisons. Analogue images can be handled within the scope of the present invention following digitization by any of numerous digitization techniques wellknown in the art, such as use of a digital camera, a scanner, among other devices and digitization techniques known in the field.
 [0041]In the comparison of two digital images, it is advantageous in some embodiments of the present invention to preprocess the images to erase any inessential difference before presenting them as input to the SSO. Such preprocessing removal of inessential differences can improve the speed to SSO processing, further enhancing the range of potential applications amenable to SSO processing.
 [0042]Also, by way of illustration and not limitation, it will be presumed in our descriptions that the images are viewed on a particular display called the reference display, and viewed at a particular viewing distance. Techniques are wellknown in the art for translating an image on a nonreference display into a digital representation as it would appear on the reference display, and for translating from an arbitrary viewing distance and angle to a standard viewing distance and angle.
 [0043]Typical inputs in the construction of a Spatial Standard Observer (SSO) are two digital images having (or scaled so as to have) the same size, called herein a test image and a reference image. G(x,y) is defined to be the grayscale of the pixel at column x and row y; G_{test}(x,y), G_{reference}(x,y) for the test and reference images respectively. We take the dimension of the image to be n_{x }pixels in the x direction (width) and n_{y }pixels in the y direction (height). Typical values are n_{x}=640 and n_{y}=480.
 [0044]Letting s_{x }and s_{y }be the viewing angles subtended by the image in the x and y directions respectively, the viewing angles s_{x}, s_{y }can be derived from the viewing distance and the image size in the plane of the display by the use of Eq. 1 twice, once to compute s_{x }and once to compute s_{y}.
$\begin{array}{cc}\mathrm{tan}\{\left(\Pi *\mathrm{size}\left(\mathrm{degrees}\right)/360\right\}=\left(0.5*\mathrm{size}\left(\mathrm{cm}\right)\right)/\mathrm{viewing}\text{\hspace{1em}}\mathrm{distance}\text{\hspace{1em}}\left(\mathrm{cm}\right)& \mathrm{Eq}.\text{\hspace{1em}}1a\\ \text{\hspace{1em}}\mathrm{size}\left(\mathrm{degrees}\right)=\frac{360}{2\pi}\frac{\mathrm{size}\left(\mathrm{cm}\right)}{\mathrm{viewing}\text{\hspace{1em}}\mathrm{distance}\text{\hspace{1em}}\left(\mathrm{cm}\right)}& \mathrm{Eq}.\text{\hspace{1em}}1b\end{array}$
Eq. 1b follows from Eq. 1a only when the ratio (size/(viewing distance)) is much less than one. But this is true in virtually all cases of practical interest so we use Eq. 1b hereinafter. Also, the designation of cm in Eq. 1a and 1b is for convenience, since it is only necessary that “size” and “viewing distance” be expressed in the same units of length.  [0045]The width and height of each pixel, p_{x }and p_{y }respectively, are given by Eq. 2 with p_{x}, p_{y }in degrees if s_{x }and s_{y }are in degrees. Typical values are s_{x}=8 deg. and s_{y}=6 deg. yielding typical values for p_{x}=p_{y}=( 1/80) deg.
$\begin{array}{cc}{p}_{x}=\frac{{s}_{x}}{{n}_{x}},{p}_{y}=\frac{{s}_{y}}{{n}_{y}}& \mathrm{Eq}.\text{\hspace{1em}}2\end{array}$  [0046]The test and reference images, G_{test}(x,y) and G_{reference}(x,y) respectively, may contain noise, or may differ in those image components having high spatial frequencies whose visibilities are not of interest for the particular image analysis under consideration. In addition, the images may be captured at a higher resolution or larger area than is necessary for the particular image analysis. For these and other reasons, it may be useful to preprocess the test and reference images to remove noise, remove high frequency components and other components not significantly affecting the visibility analysis, to reduce image resolution, and/or to crop the image to a rectangle of interest (or other convenient shape). Such operations can be performed by filtering, downsampling and cropping, pursuant to some embodiments of the present invention. Such operations are optional and, when employed, can be employed in any combination, sequence or number. That is, multiple steps of each operation can be performed whenever advantageous to do so, and the sequence of various operations or combinations can also be adjusted for the particular image processing task at hand. To be concrete in our description, we describe typical preprocessing operations, individually and in a particular sequence, understanding thereby that the present invention is not limited to the particular steps, sequence, number or type of operations described.
 [0047]It is convenient in some embodiments to prefilter the test and reference images by convolution with a prefilter function PF(x,y) pursuant to Eq. 2.1
G′(x,y)=PF(x,y){circle around (×)}G(x,y) Eq. 2.1
for G_{test}(x, y) and G_{reference}(x,y) respectively. The G′ function of Eq. 2.1, the preprocessed image, is then used in place of G in subsequent image processing, including in Eqs. 3, 4 and following.  [0048]In some embodiments of the present invention, it is convenient to use a prefilter function PF(x,y) given by Eq. 2.2.
$\begin{array}{cc}\begin{array}{c}\mathrm{PF}\left(x,y\right)=\mathrm{PF}\left(r\right)\\ =\frac{1}{{\mathrm{pscale}}^{2}}\mathrm{Exp}\left({\pi \left(\frac{r}{\mathrm{pscale}}\right)}^{2}\right)\end{array}\text{}r=\sqrt{{\left(x\text{\hspace{1em}}{p}_{x}\right)}^{2}+{\left(y\text{\hspace{1em}}{p}_{y}\right)}^{2}}& \mathrm{Eq}.\text{\hspace{1em}}2.2\end{array}$
in which pscale is a parameter, conveniently taken to be 0.125 degree in some embodiments.  [0049]The test and reference images can be downsampled by integer factors in the x and y directions {d_{x}, d_{y}} respectively, by selecting every d_{x}th column and d_{y}th row from the original image to create a new, downsampled image G″(x,y). This operation is conveniently expressed in terms of a “downsampling operator” DS as
G″(x,y)=DS(G′(x,y),d_{x} ,d _{y}) Eq. 2.3  [0050]The new dimensions of the test and reference images in the x and y directions are thus given as n_{x}′ and n_{y}′ as in Eq. 2.4.
$\begin{array}{cc}{n}_{x}^{\prime}=\mathrm{Floor}\left(\frac{{n}_{x}}{{d}_{x}}\right)\text{}{n}_{y}^{\prime}=\mathrm{Floor}\left(\frac{{n}_{y}}{{d}_{y}}\right)& \mathrm{Eq}.\text{\hspace{1em}}2.4\end{array}$
in which the function “Floor[ ]” returns the nearest integer less than or equal to its argument. Typical values for d_{x }and d_{y }are d_{x}=d_{y}=4.  [0051]Eq. 2.3 uses the preprocessed image G′ from Eq. 2.1 as the image from which the downsampled image G″ is derived. This is a particular example presented to be concrete in our description and not intending to limit the scope of the present invention. Although downsampling is almost always preceded by filtering to avoid aliasing, downsampling can be performed on an image with or without prefiltering.
 [0052]The image G, G′ or G″ can be cropped to a rectangle of interest ROI. For definiteness, we describe cropping the G″ image having dimensions n_{x}′ and n_{y}′. It is convenient to describe the ROI by the pixel coordinates of its lower left corner {x_{LL}, y_{LL}} and upper right corner {x_{UR}, y_{UR}} respectively. Cropping is conveniently performed by deleting from the image rows 1 through (y_{LL}−1) inclusive, and rows (y_{UR}+1) through n_{y}′ inclusive, as well as columns 1 through (x_{LL}−1) inclusive, and columns (x_{UR}+1) through n_{x}′ inclusive. The dimensions of the new, cropped image are thus
n″_{x} =x _{UR} −x _{LL}+1
n″ _{y} =y _{UR} −y _{LL}+1 Eq. 2.5  [0053]If the preprocessing techniques are used, singly or in combination, the resulting output images (test and reference) are considered the input images to the other image processing procedures described herein. New image dimensions (if present) should also be used.
 [0054]If a reference image is not readily available, it is convenient in some embodiments of the present invention to create one by deleting the target or structural component from a copy of the test image. If the target is confined to a local region on an otherwise uniform image with graylevel G_{0}, then it is convenient in some embodiments of the present invention to create a reference image as a uniform image having the same size as the test image with a graylevel also equal to G_{0}. Typical images are depicted as 100 in
FIG. 1 with test image 100 a and reference image 100 b. The structural component of the test image 100 a is shown adjacent to the image field merely for clarity of depiction, understanding that the images are actually superimposed.  [0055]If a reference image is not available, some embodiments of the present invention obtain a reference image by processing the test image, for example, convolving the test image with a reference filter, RF(x,y). It is advantageous in some embodiments to preprocess the test image pursuant to one or more of the preprocessing techniques described herein (or others known in the field) before application of the reference filter, that is, convolve RF with G, G′, G″ or equivalents, among others.
 [0056]In some embodiments, it is convenient to create a reference image by smoothing the test image and thereby suppress from the test image the signals whose visibility is of interest. For example, smoothing can conveniently be carried out with a Gaussian reference filter having the form given by Eq. 2.6.
$\begin{array}{cc}\begin{array}{c}\mathrm{RF}\left(x,y\right)=\mathrm{RF}\left(r\right)\\ =\frac{1}{{\mathrm{rscale}}^{2}}\mathrm{Exp}\left({\pi \left(\frac{r}{\mathrm{rscale}}\right)}^{2}\right)\end{array}\text{}r=\sqrt{{\left(x\text{\hspace{1em}}{p}_{x}\right)}^{2}+{\left(y\text{\hspace{1em}}{p}_{y}\right)}^{2}}& \mathrm{Eq}.\text{\hspace{1em}}2.6\end{array}$
“rscale” is a parameter conveniently chosen to be 2 degree.  [0057]The reference image is then created by convolving the test image with the reference filter, Eq. 2.6, either by employing conventional convolution (e.g., Eq. 5a, 5b, 5c) or, advantageously according to some embodiments of the present invention, using “confined convolution,” denoted by a “confined convolution operator” {circle around (×)}_{c}, as applied in Eq. 2.7.
G′″(x,y)=RF(x,y){circle around (×)}_{C} G″(x,y) Eq. 2.7  [0058]Eq. 2.7 depicts the example in which the preprocessed image G″″ is convolved by confined convolution to produce a reference image G′″, understanding that preprocessing the test image is optional and conventional or other forms of convolution can be employed.
 [0059]Confined convolution offers some advantages in image processing. In standard cyclic convolution, the edges of the image are considered to be connected. Thus, image content close to one edge of the image may be spread over to the opposite edge, which is sometimes called the “wraparound problem.” Confined convolution is a form of convolution which avoids the wraparound problem by, in effect, disconnecting the opposing edges of the image.
 [0060]Confined convolution makes use of a “PadConvolveCrop” (PCC) operator. The operands of the PCC operator are a general image function, I(x,y), and a kernel K(x,y) in which the kernel has k_{x }columns and k_{y }rows. The image I(x,y) is augmented or “padded” with rows and columns containing entries having a value of 0, such that the padded image has k_{x }additional columns (of all 0's) and k_{y }additional rows (of all 0's) in comparison with I(x,y). This padded I(x,y) is convolved with the kernel K(x,y). The image resulting from this convolution is then restored to the original image size by removing the added k_{y }rows and k_{x }columns. This sequence of operations defines the PCC operator operating on K and I, denoted as PCC(K(x,y),I(x,y)).
 [0061]The confined convolution of K(x,y) with I(x,y) is then given by Eq. 2.8.
$\begin{array}{cc}K\left(x,y\right){\otimes}_{C}I\left(x,y\right)=\frac{\mathrm{PCC}\left(K\left(x,y\right),I\left(x,y\right)\right)}{\mathrm{PCC}\left(\frac{K\left(x,y\right)}{\sum _{y}\sum _{x}K\left(x,y\right)},1\left(x,y\right)\right)}& \mathrm{Eq}.\text{\hspace{1em}}2.8\end{array}$
in which 1(x,y) is an image (array) all of whose entries=1 and which has the same number of rows and columns as the (unpadded) image I(x,y).  [0062]The reference and test images (optionally, following preprocessing) are converted from a grayscale format to local luminance contrast image. This conversion is depicted schematically as “Contrast” 101 in
FIG. 1 . The first step in this conversion or image transformation is the computation of a luminance image L(x,y) from the grayscales of each image, test and reference, denoted generally as G(x,y) to indicate either G_{test}(x,y) or G_{reference}(x,y) respectively. For economy of language we frequently omit subscripts “test” and “reference” using a single unsubscripted letter to indicate two equations or two variables, one each for “test” and “reference.”  [0063]This transformation from grayscale G(x,y) to a luminance image or luminance L(x,y) is advantageously performed by a gamma function “Gamma” as in Eq. 3.
L(x,y)=Gamma[G(x,y)] Eq. 3  [0064]The particular form and parameters used for the Gamma function will depend on the particular characteristics of the device displaying the test and reference images. A typical version is Eq. 4 in which the luminance L(x,y) is given by:
L(x,y)=L _{max}(G(x,y)/G _{max}) Eq. 4
in which L_{max }is the maximum possible luminance in the image, G_{max }is the corresponding maximum grayscale value. γ is the gamma exponent of the display, approximately correcting for nonlinearities in the luminance characteristics of the display. A typical value for γ is γ=2.2. Eq.s 3 and 4 are applied to both test and reference images.  [0065]A local luminance filter is employed having a luminance filter function LF(x,y). It is then convenient to introduce a local mean luminance reference image LL(x,y) obtained by the convolution of the reference luminance image L_{reference}(x,y) with the luminance filter function by Eq. 5a
LL(x,y)=LF(x,y){circle around (×)}L _{reference}(x,y) Eq. 5a
in which {circle around (×)} denotes convolution of the two functions defined in known texts in the field, for example “Fourier Analysis and Imaging” by Roger N. Bracewell, (Kluwer Academic/Plenum Publishers, 2003), pp. 174179, incorporated herein by reference. The convolution can be expressed in discrete and continuous forms as in Eq. 5b and 5c respectively.
LF(x,y){circle around (×)}L _{reference}(x,y)=∫∫LF(x−τ,y−ω)L _{reference}(x,y)dτdω Eq. 5b
where the integrals extend over the domain in which LF(τ,ω) is not zero. In discrete form the convolution is given by Eq. 5c.$\begin{array}{cc}\mathrm{LF}\left(x,y\right)\otimes {L}_{\mathrm{reference}}\left(x,y\right)=\sum _{k=0}^{{n}_{y}1}\sum _{j=0}^{{n}_{x}1}\mathrm{LF}\left(\mathrm{Mod}\left(x\text{}j,{n}_{x}\right),\mathrm{Mod}\left(y\text{}k,{n}_{y}\right)\right){L}_{\mathrm{reference}}\left(j,k\right)\text{}\mathrm{where}\text{\hspace{1em}}\mathrm{Mod}\left(a,b\right)\text{\hspace{1em}}\mathrm{is}\text{\hspace{1em}}\mathrm{the}\text{\hspace{1em}}\mathrm{remainder}\text{\hspace{1em}}\mathrm{when}\text{\hspace{1em}}a\text{\hspace{1em}}\mathrm{is}\text{\hspace{1em}}\mathrm{divided}\text{\hspace{1em}}\mathrm{by}\text{\hspace{1em}}b.& \mathrm{Eq}.\text{\hspace{1em}}5c\end{array}$  [0066]In some embodiments of the present invention, it is convenient to use the luminance filter function LF(x,y) given by Eq. 6.
$\begin{array}{cc}\begin{array}{c}\mathrm{LF}\left(x,y\right)=\mathrm{LF}\left(r\right)\\ =\frac{1}{{\mathrm{lscale}}^{2}}\mathrm{Exp}\left({\pi \left(\frac{r}{\mathrm{lscale}}\right)}^{2}\right)\end{array}\text{}r=\sqrt{{\left(x\text{\hspace{1em}}{p}_{x}\right)}^{2}+{\left(y\text{\hspace{1em}}{p}_{y}\right)}^{2}}& \mathrm{Eq}.\text{\hspace{1em}}6\end{array}$
in which 1scale is a parameter to be chosen. If 1scale→∞, this corresponds to an LL that is constant and equal to the average luminance over the image.  [0067]The average (MEAN) luminance, L_{mean }is given by a numerical average of the luminance over all pixels in the x and y directions, Eq. 7.
$\begin{array}{cc}{L}_{\mathrm{mean}}=\frac{1}{{n}_{x}{n}_{y}}\stackrel{{n}_{y}}{\sum _{y}}\stackrel{{n}_{x}}{\sum _{x}}{L}_{\mathrm{reference}}\left(x,y\right)& \mathrm{Eq}.\text{\hspace{1em}}7\end{array}$
A typical value for L_{mean }is 40 candelas per sq. meter (40 cd/m^{2}).  [0068]The contrast or contrast image of each pixel, C(x,y) is then given by Eq. 8 applied to both test and reference luminance images L_{test}(x,y) and L_{reference}(x,y) respectively.
$\begin{array}{cc}C\left(x,y\right)=\frac{L\left(x,y\right)}{L\text{\hspace{1em}}L\left(x,y\right)}1& \mathrm{Eq}.\text{\hspace{1em}}8\end{array}$  [0069]For the particular embodiments described thus far, L_{max }plays no apparent role since it appears as a multiplicative factor in both L (Eq. 4), and LL (through L_{reference }(Eq.s 4 and 5)) hence canceling from Eq. 8. (Under the typically reasonable presumption that both test and reference images have the same maximum possible luminances, L_{max}.) However, it is convenient to retain L_{max }in the equations since it simplifies the application of the equations in other embodiments of the present invention in which L_{max }and/or L_{mean }may play a role in determining parameters of the process. A typical value for L_{max }is 100 cd/m^{2}.
 [0070]Following the construction of test and reference contrast functions via Eq. 8, both test and reference images are typically passed through a Contrast Sensitivity Filter (CSF), 102 in
FIG. 1 . While various embodiments of CSF are feasible and can be used in connection with the present invention, it is advantageous in connection with some embodiments of the present invention to work in the frequency domain following application of a Discrete Fourier Transform, DFT, and its inverse DFT^{−1}. In such cases, the filtering can be described by Eq. 9 as
F(x,y)=DFT ^{−1} [CSF(u,v)*DFT[C(x,y)]] Eq. 9
in which C(x,y) is the contrast function of the image from Eq. 8 and F(x,y) is the filtered image.  [0071]The Discrete Fourier Transform and the Inverse Discrete Fourier Transform, DFT[ ] and DFT^{−1}[ ], are conventional operations in the field of digital signal processing and described in many texts, for example, the text by Bracewell, supra at pp. 167168, incorporated herein by reference.
 [0072]CSF(u,v) is the discrete version of a Contrast Sensitivity Filter in the frequency domain, and u and v are horizontal and vertical frequency indices respectively in units of cycles/width and cycles/height.
 [0073]The discrete, frequency domain version of the Contrast Sensitivity filter, CSF(u,v) is conveniently given by the product of a radial contrast sensitivity function, RCSF(u,v), and an oblique effect contrast sensitivity filter, OEF(u,v), as expressed in Eq. 10.
CSF(u,v)=RCSF(u,v)OEF(u,v) Eq. 10  [0074]In some embodiments of the present invention it is convenient to choose a radial function RCSF having the form given in Eq. 11.
$\begin{array}{cc}\begin{array}{c}\mathrm{RCSF}\text{\hspace{1em}}\left(u,v\right)=\mathrm{RCSF}\text{\hspace{1em}}\left(f\right)\\ =\mathrm{gain}\text{\hspace{1em}}\mathrm{sech}\text{\hspace{1em}}\text{\hspace{1em}}\left({\left(\frac{f}{{f}_{0}}\right)}^{p}\right)\mathrm{loss}\text{\hspace{1em}}\mathrm{sech}\text{\hspace{1em}}\left(\frac{f}{{f}_{1}}\right)\\ f=\sqrt{{\left(\frac{u}{{s}_{x}}\right)}^{2}+{\left(\frac{v}{{s}_{y}}\right)}^{2}}\end{array}& \mathrm{Eq}.\text{\hspace{1em}}11\end{array}$
in which “sech” is the hyperbolic secant function, “gain”, “loss”, f_{0}, f_{1}, and p are parameters. Typical values for these parameters are as follows:
 f_{0}=4.173
 f_{1}=1.362
 loss=0.8493
 gain=373.1
 p=0.7786.

 [0080]In some embodiments of the present invention, it is convenient to choose an oblique filter, OEF having the form given in Eq. 12.
$\begin{array}{cc}\begin{array}{c}\mathrm{OEF}\text{\hspace{1em}}\left(u,v\right)=\mathrm{OEF}\text{\hspace{1em}}\left(f,\theta \right)\\ =1\left(1\mathrm{Exp}\text{\hspace{1em}}\left(\frac{f\mathrm{corner}}{\mathrm{slope}}\right)\right)\\ {\mathrm{Sin}}^{2}\left(2\theta \right)\text{\hspace{1em}}\mathrm{if}\text{\hspace{1em}}f>\mathrm{corner}\\ =1\text{\hspace{1em}}\mathrm{if}\text{\hspace{1em}}f\le \mathrm{corner}\\ \theta =\mathrm{ArcTan}\text{\hspace{1em}}\left(\frac{u}{{s}_{x}},\frac{v}{{s}_{y}}\right)\end{array}& \mathrm{Eq}.\text{\hspace{1em}}12\end{array}$
in which “corner” and “slope” are parameters. Typical values for “corner” and “slope” are corner=3.481 and slope=13.57149.  [0081]Following processing of both the test image and the reference image by CSF, 102, the resulting filtered images are subtracted pixelbypixel, 103. The result is the difference image D(x,y) of Eq. 13.
D(x,y)=F _{test}(x,y)−F _{reference}(x,y) Eq. 13  [0082]In some embodiments of the present invention, it is advantageous to create a mask image, M(x,y), from the filtered reference image F_{reference}(x,y). In such embodiments, the absolute value of the filtered reference image is raised to a power “a”, convolved with a masking filter MF(x,y), added to the constant 1 and the b'th root of the resulting expression is computed as in Eq. 14.
$\begin{array}{cc}M\text{\hspace{1em}}(x,y\text{\hspace{1em}})={\left[1+\mathrm{MF}\text{\hspace{1em}}\left(x,y\right)\otimes {\uf603{F}_{\mathrm{reference}}\left(x,y\right)\uf604}^{a}\right]}^{\frac{1}{b}}& \mathrm{Eq}.\text{\hspace{1em}}14\end{array}$
in which the convolution operator {circle around (×)} indicates discrete convolution.  [0083]In some embodiments, it is advantageous to choose a=b=2 in Eq. 14, resulting in a mask image M(x,y) given by Eq. 15.
M(x,y)=√{square root over (1+MF(x,y){circle around (×)}F _{reference} ^{2}(x,y))} Eq. 15  [0084]Furthermore, it is advantageous in some embodiments of the present invention to choose the masking filter MF(x,y) to have the form of Eq. 16
$\begin{array}{cc}\mathrm{MF}\text{\hspace{1em}}\left(x,y\right)=\mathrm{MF}\text{\hspace{1em}}\left(r\right)=\mathrm{mgain}\text{\hspace{1em}}\mathrm{Exp}\text{\hspace{1em}}\left(\pi \text{\hspace{1em}}{\left(\frac{r}{\mathrm{mscale}}\right)}^{2}\right)& \mathrm{Eq}.\text{\hspace{1em}}16\end{array}$
in which “mgain” and “mscale” are parameters. Typical choices for mgain and mscale are mgain=0.2 and mscale=0.1.  [0085]In some embodiments of the present invention, the difference image D(x,y) is divided by the masking image to yield a masked difference image MD(x,y) according to Eq. 17.
$\begin{array}{cc}\mathrm{MD}\text{\hspace{1em}}\left(x,y\right)=\frac{D\text{\hspace{1em}}\left(x,y\right)}{M\text{\hspace{1em}}\left(x,y\right)}& \mathrm{Eq}.\text{\hspace{1em}}17\end{array}$  [0086]For those embodiments in which a mask image is not employed, the masked difference image is simply the difference image. Also, when a mask image is not employed, the subtract operation 103 can optionally precede the CSF 102.
 [0087]At this “boost” stage, 106 in
FIG. 1 , the absolute value of the masked difference image is computed, raised to a power β, and convolved with a window function W(x,y). The result of these operations is a function that is then raised to the power 1/β and multiplied by the factor (p_{x}p_{y})^{1/β} to produce a Just Noticeable Difference Image JND(x,y) as in Eq. 18. A typical value for β is β=2.408$\begin{array}{cc}\mathrm{JND}\text{\hspace{1em}}\left(x,y\right)={{\left({p}_{x}{p}_{y}\right)}^{\frac{1}{\beta}}\left[W\text{\hspace{1em}}\left(x,y\right)\otimes {\uf603\mathrm{MD}\text{\hspace{1em}}\left(x,y\right)\uf604}^{\beta}\right]}^{\frac{1}{\beta}}& \mathrm{Eq}.\text{\hspace{1em}}18\end{array}$  [0088]In some embodiments, it is advantageous to use a window function W(x,y) as given by Eq. 19
$\begin{array}{cc}W\text{\hspace{1em}}\left(x,y\right)=W\text{\hspace{1em}}\left(r\right)=\mathrm{Exp}\text{\hspace{1em}}\left(\pi \text{\hspace{1em}}{\left(\frac{r}{\mathrm{wscale}}\right)}^{2}\right)& \mathrm{Eq}.\text{\hspace{1em}}19\end{array}$
in which “wscale” is a parameter, advantageously chosen to be approximately 1.013 in some embodiments.  [0089]It is advantageous in some embodiments of the present invention to display the complete JND(x,y) image, 107 in
FIG. 1 . While optional, such a display can provide a useful visual indication of the location and magnitude of visual signals. The JND(x,y) image can be “thresholded” (setting to zero values less than a threshold, T), reduced in size (or otherwise scaled), and/or converted to a portable format to provide a compact record of the information.  [0090]The next stage in the process combines or “pools” the values of JND(x,y) of the pixels in the x and y directions to produce a single value of JND. It is convenient to use a Minkowski summation to effect this pooling with a parameter ψ as exponent, as given in Eq. 20.
$\begin{array}{cc}\mathrm{JND}={\left({p}_{x}{p}_{y}\right)}^{\frac{1}{\psi}}{\left(\sum _{y}^{{n}_{y}}\text{\hspace{1em}}\sum _{x}^{{n}_{y}}\text{\hspace{1em}}{\uf603\mathrm{JND}\text{\hspace{1em}}\left(x,y\right)\uf604}^{\psi}\right)}^{\frac{1}{\psi}}& \mathrm{Eq}.\text{\hspace{1em}}20\end{array}$
The number, JND of Eq. 20 is the desired numerical value characterizing the Spatial Standard Observer.  [0091]In some embodiments, it is advantageous to let ψ→∞, in which case Eq. 20 reduces to Eq. 21.
JND=Max[JND(x,y)] Eq. 21  [0092]In some embodiments of the present invention, it is advantageous to apply a nonlinear transformation (for example, a power function) to the JND computed from either Eq. 20 or Eq. 21. Thus, whether or not a nonlinear transformation is applied to JND, and whether or not border effects are relevant for the particular image(s) under consideration, the Spatial Standard Observer, as characterized by the value of JND, provides an effective visibility metric, able to be computed relatively rapidly.
 [0093]In some applications, the target or test image (201 in
FIG. 2 ) may be located adjacent to a border 200 of the image region 202, as depicted inFIG. 2 . If the region of the display outside the image, 200, is darker than the display, 202, for example, the dark region of a Liquid Crystal Display (LCD) panel, then the visibility of the target 201 in the region will typically be reduced. An example of this situation is depicted inFIG. 2 . Thus, it is advantageous in some embodiments of the present invention to use special techniques for the treatment of border areas in order to produce correct visibility estimates for such targets.  [0094]In some embodiments of the present invention it is advantageous to multiply the contrast images by a spatial border aperture function BA(x,y) between the Contrast and CSF steps, that is, at 120 in the process flow diagram of
FIG. 1 . The resulting Contrast Border Aperture Function, CBA(x,y) is thus
CBA(x,y)=C(x,y)BA(x,y) Eq. 22  [0095]Then CBA(x,y) is used in place of C(x,y) at the CSF step, Eq. 9.
 [0096]In some embodiments of the present invention, the border aperture function is advantageously chosen to be:
$\begin{array}{cc}\mathrm{BA}\text{\hspace{1em}}\left(x,y\right)=1\mathrm{bgain}\text{\hspace{1em}}\mathrm{Exp}\left(\pi \frac{\begin{array}{c}\mathrm{Min}\\ {\left[\begin{array}{c}\left(x1\right){p}_{x},\left(y1\right){p}_{y},\\ \left({n}_{x}x\right){p}_{x},\left({n}_{y}y\right){p}_{y\text{\hspace{1em}}}\end{array}\right]}^{2}\end{array}}{{\mathrm{bscale}}^{2}}\right)& \mathrm{Eq}.\text{\hspace{1em}}23\end{array}$
in which “bgain” and “bscale” are parameters. An example of this function is given inFIG. 3 in which the image is taken to have 240 columns (xcoordinate) and 180 rows (ycoordinate). Each pixel in this example is taken to be ( 1/60) degree in both height and width. The parameters are chosen in this example as bscale=0.5 degree and bgain=1.  [0097]The use of a border aperture function, BA(x,y) as in Eq. 23, has the advantage of simplicity, but as an approximation, it may not be as accurate as alternative methods. In other embodiments, it is advantageous for the parameters bscale and bgain to depend upon the luminance contrast between the image and the border. Typically, a margin is added to the image such that the enlarged image, image+margin, contains a portion of the border. This enlarged image is then processed as the “image” pursuant to the image processing techniques described herein, typically including the masking component of the processing, 105. The presence of a portion of the border in the enlarged image will tend to produce the appropriate masking effect, tending to reduce visibility of targets or portions of targets near the border.
 [0098]There are various ways the use of an enlarged image can be implemented to treat border effects. For example, it is convenient to take the width of the border region to be Round[2*mscale/p_{x}] and the height to be Round[2*mscale/p_{y}], in which mscale is the masking parameter (Eq. 16). “Round[ ]” is a function that generates as the value of the function that integer nearest to the value of the function's argument. The dimensions of the enlarged image are then given by Eq. 24 as:
width=n _{x}+Round[2*mscale/p _{x}]
height=n _{y}+Round[2*mscale/p _{y}] Eq. 24  [0099]An advantage of treating border effects with an enlarged image is that it more correctly deals with the dependence of the border masking effect upon the luminance contrast between the border and the (original, unenlarged) image. A possible disadvantage is that this approach requires somewhat more processing to include the masking step.
 [0100]JND from Eq. 20 (or Eq. 21 for ψ→∞) relates to the percentage of human observers who will notice a difference. For example, images leading to JND having a value around 1 will typically present noticeable differences to about 75% of typical human observers. Images resulting in larger JND values will present noticeable difference to a correspondingly larger percentage of typical human observers, although the precise functional relationship between JND and the percentage of viewers observing differences may not be readily known.
 [0101]It is advantageous in some embodiments of the present invention to use JND as a measure of different levels of perceptual intensity. That is, larger JND values indicate that a larger percentage of observers will notice a difference. But also larger values of JND typically indicate that a given observer will be more likely to observe more detailed differences. By way of illustration and not limitation, we consider the example of observing a scene through some form of optical instrument, such as a remote viewing device, night vision goggles, among others. A given observer may require an image value of JND_{1 }in order to conclude that some object is present other than natural background. However a value of JND_{2}>JND_{1 }would be required for the observer to conclude that the object is a military vehicle. And a value of JND_{3}>JND_{2 }would be required to conclude that it is a hostile military vehicle. Thus, JND values as determined by the SSO can be a useful measure of not only minimal levels of visibility but, when more stringently applied, also estimate the probable level of perceptual information obtainable from a given image.
 [0102]
FIG. 4 depicts an illustrative computer system 250 that utilizes the teachings of the present invention. The computer system 250 comprises a processor 252, a display 254, input interfaces 256, communications interface 258, memory 260, and output interfaces 262, all conventionally coupled by one or more busses 264. The input interfaces 256 comprise a keyboard 266, mouse, trackball or similar device 268, as well as massstorage input devices such as CDs, DVDs, magnetic discs of various designs among others. The output interface 262 is a printer 272. The communications interface 258 is a network interface card (NIC) that allows the computer 250 to communicate via a network, such as the Internet. Image acquisition/generation devices 274 provide the images 100 for the generation of the SSO and are also coupled to the processor 252. The units 274 can supply either stored or realtime input data, or both.  [0103]The memory 260 typically comprises different modalities, illustratively semiconductor memory, such as random access memory (RAM), and disk drives. Depending on the embodiment, the memory 260 typically includes an operating system, 280. The operating system 280 may be implemented by any conventional operating system such as UNIX®, WINDOWS®, and LINUX®, among others.
 [0104]Although various embodiments which incorporate the teachings of the present invention have been shown and described in detail herein, those skilled in the art can readily devise many other varied embodiments that still incorporate these teachings.
Claims (38)
1. A method of producing a visibility metric from a digital reference image and a digital test image, the method comprising:
a) producing a reference luminance image from the reference image and a test luminance image from the test image;
b) producing a local mean luminance reference image as a convolution of the reference luminance image and a luminance filter function;
c) producing a test contrast image as at least one of the following: (c1) a mathematical combination of the test luminance image and the local mean luminance reference image and (c2) a mathematical combination of the test luminance image, the local mean luminance reference image and a border aperture function and (c3) a mathematical combination of the test luminance image, the local mean luminance reference image and an image of a border surrounding the reference image;
d) producing a reference contrast image as at least one of the following: (d1) a mathematical combination of the reference luminance image and the local mean luminance reference image and (d2) a mathematical combination of the reference luminance image, the local mean luminance reference image and the border aperture function;
e) applying a contrast sensitivity filter to the test contrast image to produce a filtered test image;
f) applying the contrast sensitivity filter to the reference contrast image to produce a filtered reference image;
g) providing a difference image by at least one of the following two processes: (g1) subtracting the filtered reference image from the filtered test image to produce a difference image, and (g2) producing a mask image as a mathematical combination of the filtered reference image with a masking filter, and producing a difference image as a ratio of the difference image and the mask image;
h) producing a just noticeable difference image as a mathematical combination of the difference image with a window function; and
i) pooling the just noticeable difference image to produce a visibility metric.
2. (canceled)
3. (canceled)
4. A method as in claim 1 , further comprising:
providing an enlarged reference luminance image by including with the reference image a selected portion of said border surrounding said reference luminance image.
5. A method as in claim 1 further comprising preprocessing at least one of said digital reference image and said digital test image immediately preceding performance of said process (a).
6. A method as in claim 5 , wherein said preprocessing is performed by prefiltering.
7. A method as in claim 6 , wherein said prefiltering is performed by convolution with a prefiltering function.
8. A method as in claim 7 wherein said prefiltering function is a Gaussian function.
9. A method as in claim 5 wherein said preprocessing is performed by downsampling.
10. A method as in claim 5 wherein said preprocessing is performed by cropping.
11. An article of manufacture comprising a computer usable medium embodying one or more instructions executable by a computer for performing a method of producing a visibility metric, the instructions comprising:
a) producing a reference luminance image from the reference image and a test luminance image from the test image;
b) producing a local mean luminance reference image as a convolution of the reference luminance image and a luminance filter function;
c) producing a test contrast image as at least one of the following: (c1) a mathematical combination of the test luminance image and the local mean luminance reference image and (c2) a mathematical combination of the test luminance image, the local mean luminance reference image and a border aperture function and (c3) a mathematical combination of the test luminance image, the local mean luminance reference image and an image of a border surrounding said reference image;
d) producing a reference contrast image as at least one of the following: (d1) a mathematical combination of the reference luminance image and the local mean luminance reference image and (d2a mathematical combination of the reference luminance image, the local mean luminance reference image and the border aperture function;
e) applying a contrast sensitivity filter to the test contrast image to produce a filtered test image;
f) applying the contrast sensitivity filter to the reference contrast image to produce a filtered reference image;
g) providing a difference image by at least one of the following two processes: (g1) subtracting the filtered reference image from the filtered test image to produce a difference image, and (g2) producing a mask image as a mathematical combination of the filtered reference image with a masking filter, and producing a difference image as a ratio of the difference image and the mask image;
h) producing a just noticeable difference image as a mathematical combination of the difference image with a window function; and
i) pooling the just noticeable difference image to produce a visibility metric.
12. (canceled)
13. (canceled)
14. An article of manufacture as in claim 11 , wherein said instructions further comprise:
providing an enlarged reference luminance image by including with the reference image a selected portion of said border surrounding said reference luminance image.
15. An article of manufacture as claim 11 , wherein said instructions further comprise preprocessing at least one of said digital reference image and said digital test image immediately preceding performance of said st
16. An article of manufacture as in claim 15 wherein said preprocessing is performed by prefiltering.
17. An article of manufacture as in claim 16 wherein said prefiltering is performed by convolution with a prefiltering function.
18. An article of manufacture as in claim 17 wherein said prefiltering function is a Gaussian function.
19. An article of manufacture as in claim 15 wherein said preprocessing is performed by downsampling.
20. An article of manufacture as in claim 15 wherein said preprocessing is performed by cropping.
21. A spatial standard observer apparatus for producing a visibility metric, the apparatus comprising:
a computer that is programmed for:
i) receiving a digital reference image and a digital test image;
ii) producing a reference luminance image from the reference image and a test luminance image from the test image;
iii) producing a local mean luminance reference image as a convolution of the reference luminance image and a luminance filter function;
iv) producing a test contrast image as at least one of the following: (iv1) a mathematical combination of the test luminance image and the local mean luminance reference image and (iv2) a mathematical combination of the test luminance image, the local mean luminance reference image and a border aperture function and (iv3) a mathematical combination of the test luminance image, the local mean luminance reference image and an image of a border surrounding the reference image;
v) producing a reference contrast image as at least one of the following: (v1) a mathematical combination of the reference luminance image and the local mean luminance reference image and (v2) a mathematical combination of the reference luminance image, the local mean luminance reference image and the border aperture function;
vi) applying a contrast sensitivity filter to the test contrast image to produce a filtered test image;
vii) applying said contrast sensitivity filter to the reference contrast image to produce a filtered reference image; and,
viii) providing a difference image by at least one of the following two processes; (viii1) subtracting the filtered reference image from the filtered test image to produce a difference image, and (viii2) producing mask image as a mathematical combination of the filtered reference image with a masking filter, and producing a difference image as a ratio of the difference image and the mask image; and,
ix) producing a just noticeable difference image as a mathematical combination of the difference image with a window function; and,
x) pooling the just noticeable difference image to produce a visibility metric.
22. (canceled)
23. (canceled)
24. Apparatus as in claim 21 , wherein said computer is further programmed for
providing an enlarged reference luminance image by including with the reference image a selected portion of said border surrounding said reference luminance image; and,
25. Apparatus as in claim 21 , wherein said computer is further programmed for said preprocessing of said digital reference image and said digital test image immediately preceding performance of said process (a).
26. Apparatus as in claim 25 wherein said preprocessing is performance of prefiltering.
27. Apparatus as in claim 26 wherein said prefiltering is performed by convolution with a prefiltering function.
28. Apparatus as in claim 27 wherein said prefiltering function is a Gaussian function.
29. Apparatus as in claim 25 wherein said preprocessing is performed by downsampling.
30. Apparatus as in claim 25 wherein said preprocessing is performed by cropping.
31. A method of forming a digital reference image from a digital test image, the method comprising:
a) performing a convolution of the test image with a reference filter, to thereby form the digital reference image.
32. A method as in claim 31 , further comprising choosing said reference filter to be a smoothing filter.
33. A method as in claim 32 further comprising choosing said smoothing filter to be a Gaussian function.
34. A method as in claim 31 further comprising choosing said convolution to be confined convolution.
35. A method of reducing wraparound in a digital image processing process that includes one or more convolution steps, the method comprising:
a) producing a reference luminance image from the reference image and a test luminance image from the test image;
b) producing a local mean luminance reference image as convolution of the reference luminance image and a luminance filter function;
c) producing a test contrast image as at least one of the following: (c1) a mathematical combination of the test luminance image and the local mean luminance reference image and (c2) a mathematical combination of the test luminance image, the local mean luminance reference image and a border aperture function and (c3) a mathematical combination of the test luminance image, the local mean luminance reference image and an image of a border surrounding the reference image;
d) producing a reference contrast image as at least one of the following: (d1) a mathematical combination of the reference luminance image and the local mean luminance reference image and (d2) a mathematical combination of the reference luminance image, the local mean luminance reference image and the border aperture function;
e) applying a contrast sensitivity filter to the test contrast image to produce a filtered test image;
f) applying the contrast sensitivity filter to the reference contrast image to produce a filtered reference image;
g) providing a difference image by at least one of the following two processes: (g1) subtracting the filtered reference image from the filtered test image to produce a difference image, and (g2) producing a ask image as a mathematical combination of the filtered reference image with a masking filter, and producing a difference image as a ratio of the difference image and the mask image;
h) producing a just noticeable difference image as a mathematical combination of the difference image with a window function; and
i) pooling the just noticeable difference image to produce a visibility metric,
wherein the convolution operation in at least one of process (b) and process (h) is performed using confined convolution.
36. A method of reducing wraparound in a digital imaging process that includes at least one image convolution step, the method comprising:
a) producing a reference luminance image from a reference image and a test luminance image from the test image;
b) producing a local mean luminance reference image as a convolution of the reference luminance image and a luminance filter function;
c) producing a test contrast image from the test luminance image and the local mean luminance reference image;
d) producing a reference contrast image from the reference luminance image and the local mean luminance reference image;
e) applying a contrast sensitivity filter to the test contrast image to produce a filtered test image;
f) applying the contrast sensitivity filter to the reference contrast image to produce a filtered reference image;
g) subtracting the filtered reference image from the filtered test image to produce a difference image;
h) producing a just noticeable difference image as a mathematical combination of the difference image with a window function, wherein the convolution process in at leat5 one of processes (b) and (h) includes further convolution with a reference filter function RF(x,y) that is substantially zero except for coordinates x and y lying in respective small coordinate ranges;
i) pooling said just noticeable difference image to produce a visibility metric.
37. (canceled)
38. (canceled)
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

US11045041 US7783130B2 (en)  20050124  20050124  Spatial standard observer 
Applications Claiming Priority (3)
Application Number  Priority Date  Filing Date  Title 

US11045041 US7783130B2 (en)  20050124  20050124  Spatial standard observer 
PCT/US2006/002662 WO2006079115A3 (en)  20050124  20060124  Spatial standard observer 
US12807375 US8139892B2 (en)  20050124  20100809  Spatial standard observer 
Related Child Applications (1)
Application Number  Title  Priority Date  Filing Date 

US12807375 Continuation US8139892B2 (en)  20050124  20100809  Spatial standard observer 
Publications (2)
Publication Number  Publication Date 

US20060165311A1 true true US20060165311A1 (en)  20060727 
US7783130B2 US7783130B2 (en)  20100824 
Family
ID=36693029
Family Applications (2)
Application Number  Title  Priority Date  Filing Date 

US11045041 Active 20280326 US7783130B2 (en)  20050124  20050124  Spatial standard observer 
US12807375 Active US8139892B2 (en)  20050124  20100809  Spatial standard observer 
Family Applications After (1)
Application Number  Title  Priority Date  Filing Date 

US12807375 Active US8139892B2 (en)  20050124  20100809  Spatial standard observer 
Country Status (2)
Country  Link 

US (2)  US7783130B2 (en) 
WO (1)  WO2006079115A3 (en) 
Cited By (23)
Publication number  Priority date  Publication date  Assignee  Title 

US20060064209A1 (en) *  20040920  20060323  KyungSoo Park  Car display device for use in control of playback device 
US20080012856A1 (en) *  20060714  20080117  Daphne Yu  Perceptionbased quality metrics for volume rendering 
US20080063254A1 (en) *  20060907  20080313  Kabushiki Kaisha Toshiba  Unevenness inspection method, method for manufacturing display panel, and unevenness inspection apparatus 
US20080123944A1 (en) *  20061123  20080529  Samsung Electronics Co., Ltd.  Apparatus and method for color reproduction 
US20100054623A1 (en) *  20080901  20100304  Kabushiki Kaisha Toshiba  Image display apparatus and method 
US20100053204A1 (en) *  20080826  20100304  Lg Display Co., Ltd.  Method for setting compensation region for irregular deffect region in manage display device 
US20100271461A1 (en) *  20090422  20101028  Sony Corporation  Transmitting apparatus, stereoscopic image data transmitting method, receiving apparatus, and stereoscopic image data receiving method 
US8279204B1 (en) *  20051222  20121002  The Mathworks, Inc.  Viewer for multidimensional data from a test environment 
CN102881010A (en) *  20120828  20130116  北京理工大学  Method for evaluating perception sharpness of fused image based on human visual characteristics 
US8402317B1 (en)  20051222  20130319  The Math Works, Inc.  Viewing multidimensional metric data from multiple test cases 
US8559511B2 (en)  20100330  20131015  Hong Kong Applied Science and Technology Research Institute Company Limited  Method and apparatus for video coding by ABTbased just noticeable difference model 
CN103955689A (en) *  20140414  20140730  杭州电子科技大学  Image vision areaofinterest extraction method through frequency screening 
US20140254897A1 (en) *  20130306  20140911  Tektronix, Inc.  Design verification and diagnostics for image devices 
US20140328406A1 (en) *  20130501  20141106  Raymond John Westwater  Method and Apparatus to Perform Optimal VisuallyWeighed Quantization of TimeVarying Visual Sequences in Transform Space 
CN104657996A (en) *  20150226  20150527  西安交通大学  Nonlinear normalization based IQA (image quality assessment) method of LaplaceGaussian signal 
US9380186B2 (en)  20120824  20160628  Digimarc Corporation  Data hiding for spot colors in product packaging 
US9396509B1 (en)  20111030  20160719  Digimarc Corporation  Closed form noniterative watermark embedding 
US9401001B2 (en)  20140102  20160726  Digimarc Corporation  Fullcolor visibility model using CSF which varies spatially with local luminance 
US9449357B1 (en)  20120824  20160920  Digimarc Corporation  Geometric enumerated watermark embedding for spot colors 
US9565335B2 (en)  20140102  20170207  Digimarc Corporation  Full color visibility model using CSF which varies spatially with local luminance 
US9635378B2 (en)  20150320  20170425  Digimarc Corporation  Sparse modulation for robust signaling and synchronization 
US9747656B2 (en)  20150122  20170829  Digimarc Corporation  Differential modulation for robust signaling and synchronization 
US9892478B2 (en)  20150306  20180213  Digimarc Corporation  Digital watermarking applications 
Families Citing this family (2)
Publication number  Priority date  Publication date  Assignee  Title 

US8675922B1 (en)  20110524  20140318  The United States of America as represented by the Administrator of the National Aeronautics & Space Administration (NASA)  Visible motion blur 
CN103325084B (en) *  20130627  20160316  宁波大学  Based on just noticeable distortion combined stereoscopic image fragile watermarking method 
Citations (7)
Publication number  Priority date  Publication date  Assignee  Title 

US5694491A (en) *  19960329  19971202  David Sarnoff Research Center, Inc.  Methods and apparatus for assessing the visibility of differences between two image sequences 
US5719966A (en) *  19960329  19980217  David Sarnoff Research Center, Inc.  Apparatus for assessing the visiblity of differences between two image sequences 
US5974159A (en) *  19960329  19991026  Sarnoff Corporation  Method and apparatus for assessing the visibility of differences between two image sequences 
US6148117A (en) *  19961227  20001114  HewlettPackard Company  Image processing system with alterable local convolution kernel 
US20020031277A1 (en) *  19970404  20020314  Jeffrey Lubin  Method and apparatus for assessing the visibility of differences between two signal sequences 
US6799515B2 (en) *  20020311  20041005  Jerry K Lynn  Luminous double faced picture display 
US6977664B1 (en) *  19990924  20051220  Nippon Telegraph And Telephone Corporation  Method for separating background sprite and foreground object and method for extracting segmentation mask and the apparatus 
Family Cites Families (6)
Publication number  Priority date  Publication date  Assignee  Title 

US5781665A (en) *  19950828  19980714  Pitney Bowes Inc.  Apparatus and method for cropping an image 
US7016045B2 (en) *  19990312  20060321  Regents Of The University Of Minnesota  Video camerabased visibility measurement system 
DE10034461A1 (en)  20000715  20020131  Bosch Gmbh Robert  A method for determining the visibility distance 
US6983083B2 (en)  20010213  20060103  Eastman Kodak Company  Image specific perceived overall contrast prediction 
US6816627B2 (en) *  20010412  20041109  Lockheed Martin Corporation  System for morphological image fusion and change detection 
EP1611738A2 (en)  20030327  20060104  Sergei Startchik  Method for estimating logo visibility and exposure in video 
Patent Citations (7)
Publication number  Priority date  Publication date  Assignee  Title 

US5694491A (en) *  19960329  19971202  David Sarnoff Research Center, Inc.  Methods and apparatus for assessing the visibility of differences between two image sequences 
US5719966A (en) *  19960329  19980217  David Sarnoff Research Center, Inc.  Apparatus for assessing the visiblity of differences between two image sequences 
US5974159A (en) *  19960329  19991026  Sarnoff Corporation  Method and apparatus for assessing the visibility of differences between two image sequences 
US6148117A (en) *  19961227  20001114  HewlettPackard Company  Image processing system with alterable local convolution kernel 
US20020031277A1 (en) *  19970404  20020314  Jeffrey Lubin  Method and apparatus for assessing the visibility of differences between two signal sequences 
US6977664B1 (en) *  19990924  20051220  Nippon Telegraph And Telephone Corporation  Method for separating background sprite and foreground object and method for extracting segmentation mask and the apparatus 
US6799515B2 (en) *  20020311  20041005  Jerry K Lynn  Luminous double faced picture display 
Cited By (33)
Publication number  Priority date  Publication date  Assignee  Title 

US20060064209A1 (en) *  20040920  20060323  KyungSoo Park  Car display device for use in control of playback device 
US8279204B1 (en) *  20051222  20121002  The Mathworks, Inc.  Viewer for multidimensional data from a test environment 
US8762784B1 (en)  20051222  20140624  The Mathworks, Inc.  Viewing multidimensional metric data from multiple test cases 
US8402317B1 (en)  20051222  20130319  The Math Works, Inc.  Viewing multidimensional metric data from multiple test cases 
US8780098B1 (en)  20051222  20140715  The Mathworks, Inc.  Viewer for multidimensional data from a test environment 
US20080012856A1 (en) *  20060714  20080117  Daphne Yu  Perceptionbased quality metrics for volume rendering 
US8320658B2 (en) *  20060907  20121127  Kabushiki Kaisha Toshiba  Unevenness inspection method, method for manufacturing display panel, and unevenness inspection apparatus 
US20080063254A1 (en) *  20060907  20080313  Kabushiki Kaisha Toshiba  Unevenness inspection method, method for manufacturing display panel, and unevenness inspection apparatus 
US20080123944A1 (en) *  20061123  20080529  Samsung Electronics Co., Ltd.  Apparatus and method for color reproduction 
US8411936B2 (en) *  20061123  20130402  Samsung Electronics Co., Ltd.  Apparatus and method for color reproduction 
US8085278B2 (en) *  20080826  20111227  Lg Display Co., Ltd.  Method for setting compensation region for irregular defect region in manage display device 
KR101319341B1 (en)  20080826  20131016  엘지디스플레이 주식회사  Method of generating compensation region for compensating defect of image display device 
US20100053204A1 (en) *  20080826  20100304  Lg Display Co., Ltd.  Method for setting compensation region for irregular deffect region in manage display device 
US20100054623A1 (en) *  20080901  20100304  Kabushiki Kaisha Toshiba  Image display apparatus and method 
US8810563B2 (en) *  20090422  20140819  Sony Corporation  Transmitting apparatus, stereoscopic image data transmitting method, receiving apparatus, and stereoscopic image data receiving method 
US20100271461A1 (en) *  20090422  20101028  Sony Corporation  Transmitting apparatus, stereoscopic image data transmitting method, receiving apparatus, and stereoscopic image data receiving method 
US8559511B2 (en)  20100330  20131015  Hong Kong Applied Science and Technology Research Institute Company Limited  Method and apparatus for video coding by ABTbased just noticeable difference model 
US9396509B1 (en)  20111030  20160719  Digimarc Corporation  Closed form noniterative watermark embedding 
US9864919B2 (en)  20120824  20180109  Digimarc Corporation  Including information signals in product packaging and other printer media 
US9710871B2 (en)  20120824  20170718  Digimarc Corporation  Geometric enumerated watermark embedding for spot colors 
US9449357B1 (en)  20120824  20160920  Digimarc Corporation  Geometric enumerated watermark embedding for spot colors 
US9380186B2 (en)  20120824  20160628  Digimarc Corporation  Data hiding for spot colors in product packaging 
CN102881010A (en) *  20120828  20130116  北京理工大学  Method for evaluating perception sharpness of fused image based on human visual characteristics 
US20140254897A1 (en) *  20130306  20140911  Tektronix, Inc.  Design verification and diagnostics for image devices 
US20140328406A1 (en) *  20130501  20141106  Raymond John Westwater  Method and Apparatus to Perform Optimal VisuallyWeighed Quantization of TimeVarying Visual Sequences in Transform Space 
US9401001B2 (en)  20140102  20160726  Digimarc Corporation  Fullcolor visibility model using CSF which varies spatially with local luminance 
US9565335B2 (en)  20140102  20170207  Digimarc Corporation  Full color visibility model using CSF which varies spatially with local luminance 
US9805435B2 (en)  20140102  20171031  Digimarc Corporation  Full color visibility model using CSF which varies spatially with local luminance 
CN103955689A (en) *  20140414  20140730  杭州电子科技大学  Image vision areaofinterest extraction method through frequency screening 
US9747656B2 (en)  20150122  20170829  Digimarc Corporation  Differential modulation for robust signaling and synchronization 
CN104657996A (en) *  20150226  20150527  西安交通大学  Nonlinear normalization based IQA (image quality assessment) method of LaplaceGaussian signal 
US9892478B2 (en)  20150306  20180213  Digimarc Corporation  Digital watermarking applications 
US9635378B2 (en)  20150320  20170425  Digimarc Corporation  Sparse modulation for robust signaling and synchronization 
Also Published As
Publication number  Publication date  Type 

US7783130B2 (en)  20100824  grant 
WO2006079115A2 (en)  20060727  application 
WO2006079115A3 (en)  20090507  application 
US8139892B2 (en)  20120320  grant 
US20100329585A1 (en)  20101230  application 
Similar Documents
Publication  Publication Date  Title 

Aydin et al.  Dynamic range independent image quality assessment  
Kovesi  Image features from phase congruency  
Wang et al.  Local phase coherence and the perception of blur  
US6628842B1 (en)  Image processing method and apparatus  
US6731821B1 (en)  Method for enhancing compressibility and visual quality of scanned document images  
Comer et al.  Morphological operations for color image processing  
US6285798B1 (en)  Automatic tone adjustment by contrast gaincontrol on edges  
US6741753B1 (en)  Method and system of local color correction using background liminance masking  
US6233060B1 (en)  Reduction of moiré in screened images using hierarchical edge detection and adaptivelength averaging filters  
US6993204B1 (en)  High speed signal enhancement using pixons  
US5905819A (en)  Method and apparatus for hiding one image or pattern within another  
Kim et al.  An advanced contrast enhancement using partially overlapped subblock histogram equalization  
US20050089239A1 (en)  Method for improving digital images and an image sensor for sensing the same  
US7715657B2 (en)  Method, device and program for detecting perceptual features of a larger image and incorporating information of the detected perceptual features into a smaller preview image  
US20110096987A1 (en)  Noise reduced color image using panchromatic image  
Nill et al.  Objective image quality measure derived from digital image power spectra  
US20090027558A1 (en)  Apparatus and Method for Rendering High Dynamic Range Images for Standard Dynamic Range Display  
US20040136603A1 (en)  Enhanced wide dynamic range in imaging  
US20020186309A1 (en)  Bilateral filtering in a demosaicing process  
US20030189579A1 (en)  Adaptive enlarging and/or sharpening of a digital image  
US6094508A (en)  Perceptual thresholding for gradientbased local edge detection  
US20050254722A1 (en)  System and method for compressing the dynamic range of an image  
US6792162B1 (en)  Method and apparatus to automatically enhance the quality of digital images by measuring grain trace magnitudes  
US7613363B2 (en)  Image superresolution through edge extraction and contrast enhancement  
US6654013B1 (en)  Apparatus for and method of enhancing shape perception with parametric texture maps 
Legal Events
Date  Code  Title  Description 

AS  Assignment 
Owner name: USA AS REPRESENTED BY THE ADMINISTRATOR OF THE NAS Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:WATSON, ANDREW B.;REEL/FRAME:016852/0410 Effective date: 20050627 

FPAY  Fee payment 
Year of fee payment: 4 